Thread: Regression: estimating the error term variance w/o fitting a regression line

There's a question in my regression class I'm having trouble with. We are basically given 6 points (5,y1), (5,y2), (10,y3), (10, y4), (15,y5), (15,y6)


and we are asked how it is possible to estimate the error term variance without fitting a regression line.




I'm thinking maybe the solution has something to do the with formula for sample variance? But I really don't know. Can anyone help?

Hey MN1987.

Hint: Think about the regression model in terms of the distribution and then consider how to find the variance of the y term in relation to that of the error term. An example model would be Y = B0 + B1X + e